The Effective Statistician - in association with PSI Linear Mixed Models - A Refresher And Introduction
May 6, 2024
This podcast explores the importance and evolution of linear mixed models, emphasizing their role in statistical analysis. It discusses modeling repeated measurements with random effects, sample size influences, Bayesian models, and handling missing values. Additionally, it covers modeling assumptions, software preferences, and resources for effective statistical modeling, highlighting the significance of rational assumptions and the preference for using SAS in modeling.
AI Snips
Chapters
Transcript
Linear Models: Linear in Coefficients
- Linear models are linear in the coefficients, not necessarily in the variables.
- You can include quadratic, logarithmic, or interaction terms and still have a linear model.
Role of Random Effects
- Random effects account for correlations within clusters or repeated measures.
- They add variability centered around zero, capturing dependencies not explained by fixed effects.
Covariance Structure Choices
- Covariance structures can vary from unstructured to autoregressive or compound symmetry.
- More flexible structures need more parameters, reducing degrees of freedom and increasing estimation challenges.